[CP2K-user] [CP2K:17198] Transition metal oxide SCF Convergence

[Nicholas Winner]

Some suggestions for what to try:

(1) Precision
 Your value of EPS_PGF_ORB is too large and should be substantially
decreased. I think the default value is too large for a lot of systems.
Most numerical issues stem from the overlap matrix precision in my opinion.
Set it to something like 1e-16 and see what happens. That might be too
fine, but its not very much extra computational cost and it can make a huge
difference.

(2) Try these settings

&OT
 algorithm irac # do not enforce strict orthogonality
 minimizer cg # safer than DIIS, though often its too safe and makes the
calculation slow.
 linesearch 2pnt # 3pnt is not compatible with irac+rotation
 rotation true # allow for subspace rotations (fractional occupations)
 preconditioner full_single_inverse # full_all not compatible with rotations
&end

If that doesn't work, decrease STEPSIZE in &OT to be 0.05

Hope that helps,
Nick

On Tue, Jun 21, 2022 at 8:45 AM 'Börries von Seggern' via cp2k <
cp2k at googlegroups.com> wrote:

> Dear Cp2k Forum,
> For a while now, I have been trying to calculate XAS Spectra of transition
> metal oxides using the LR-TDDFT method. Unfortunately, I wasn`t able to
> make the preceeding Hybrid-DFT  SCF calculation converge for open-shell
> systems (CuO, CoO, NiO), while other closed shell systems (MgO, ZnO) show
> rapid convergence quite easily. So far, I've tried using the OT methods,
> which do seem to converge until reaching a sort of plateau, where the
> energy starts increasing in miniscule amounts. Further, Iǜe been trying to
> use diaganolization, however these calculations seem to oscillate even when
> including very little amounts (1e-2) of the new density.
> Concerning the OT methods, I ve tried both DIIS and CG (with 2pnt and 3pnt
> linesearch), however, all calculations show the same behaviour.
> Could anyone please help me with these calculations?
>
> Best regards,
>
> Börries
>
> PS: Please find an example input, output and structure files attached, I
> am using cp2k - 8.2. As even this calculation using the CG optimizer with
> linesearch 3pnt reached the plateau and was very expensive, I unfortunately
> cancelled the calculation. (But could restart it if necessary)
>
> --
> You received this message because you are subscribed to the Google Groups
> "cp2k" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to cp2k+unsubscribe at googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/cp2k/149a0cc5-5ea3-4a94-b3ea-b893b725b9ddn%40googlegroups.com
> <https://groups.google.com/d/msgid/cp2k/149a0cc5-5ea3-4a94-b3ea-b893b725b9ddn%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>

-- 
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAE8AJWGYVOg73%3DEyF5-AOacrNruq4Rwv9rD%2BqMvkKiZModm7vw%40mail.gmail.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20220621/ec9e2336/attachment.htm>